Inspiration
Software development has a lot in common with Formula 1. In both worlds, speed is everything, but reliability is the difference between winning and crashing.
When we saw the Codegeist 2025: Atlassian Williams Racing Edition challenge, we asked ourselves: Why is incident management still so slow? When a Formula 1 car boxes, the crew knows exactly what to do. But when a developer hits a "Blocker" bug, they often scramble—searching for logs, guessing which commit broke the build, and digging through outdated docs.
We wanted to build the "Race Strategist" for DevOps. We were inspired by the precision of the Williams Racing pit crew and wanted to bring that same efficiency to Jira using Atlassian's newest AI capabilities.
What it does
PitCrew AI is an intelligent "Pit Wall" embedded directly inside Jira. It accelerates the "Inner Dev Loop" by automating the triage of critical bugs.
- The Pit Wall (UI): When a Jira issue is flagged as "High Priority" or "Blocked," our Forge UI Panel activates, displaying a "Red Flag" status and visual telemetry of the issue.
- The Strategist (Rovo Agent): Developers can summon the PitCrew Strategist (a specialized Rovo Agent) directly in the side chat.
- Instant Analysis: The Agent doesn't just chat; it performs actions. It scans the connected Bitbucket repository for the last 5 commits, correlates them with the crash logs, and searches Confluence for relevant runbooks.
- The Recovery Plan: The Agent returns a specific, actionable strategy (e.g., "Rollback commit a1b2c (Auth Module) by DevDriver1"), effectively acting as a senior engineer who unblocks the team instantly.
How we built it
- We built PitCrew AI entirely on the Atlassian Forge platform to ensure security and native performance.
- Backend: We used Forge Resolvers (Node.js) to handle the logic between Jira and our internal "Telemetry" engine.
- AI Logic: We defined a custom rovo:agent in the manifest.yml with a specialized "System Prompt" designed to act like an F1 Strategist.
- Tools: We built a custom Forge action (Tool) that allows the agent to fetch "Context" (simulated Bitbucket commit history and log data) so it can make grounded decisions rather than hallucinating.
- Frontend: We used the Forge UI Kit to build the "Pit Wall" panel, ensuring it looks and feels like a native part of Jira.
Challenges we ran into
- Thinking in "Agents" vs. "Chatbots": It was challenging to stop thinking about simple Q&A and start thinking about Agency. We had to design the action carefully so the AI knew when to fetch commit data and how to interpret it.
- Prompt Engineering for Rovo: Getting the Agent to adopt the "Williams Racing" persona (using terms like "Box Box" or "Green Flag") while remaining technically accurate required several iterations of prompt tuning.
Accomplishments that we're proud of
- Seamless Integration: We successfully connected the "Chat" experience with the "Visual" experience on the Jira card.
- Built with Rovo Dev: We successfully used the Rovo Dev CLI to scaffold the initial project and generate the boilerplate code, which saved us hours of setup time.
- The "Wow" Factor: We believe we managed to make the boring task of "bug triage" feel exciting and urgent, fitting the Williams Racing theme perfectly.
What we learned
- The Power of Rovo: We learned that Rovo Agents are incredibly powerful when given access to specific tools. The ability to define an action in the manifest opens up endless possibilities for automation.
- Forge is Fast: Using the Forge CLI and Rovo Dev together allowed us to go from "Idea" to "Working App" in record time.
What's next for PitCrew AI
- Real Telemetry: Connecting to real-time observability tools (like Datadog or Prometheus) via Forge External Auth.
- Slack Integration: Allowing the "Strategist" to post the "Pit Stop Plan" directly into the team's Slack channel for faster visibility.
Log in or sign up for Devpost to join the conversation.